the star pixel detector - agenda (indico) · 2016-09-05 · extend the measurement capabilities in...
TRANSCRIPT
G. Contin | [email protected]
The STAR Pixel detector
Giacomo Contin for the STAR Collaboration Lawrence Berkeley National Laboratory
PIXEL 2016 - 8th International Workshop on Semiconductor Pixel Detectors for Particles and Imaging Sestri Levante, 5-9 September 2016
G. Contin | [email protected]
Outline
A MAPS-based detector Construction Operations, Performance,
Lessons Learned Conclusions
PIXEL2016 - Sestri Levante 2 9/5/2016
G. Contin | [email protected]
SSD r = 22
IST r = 14
PXL r2 = 8
r1 = 2.8
The STAR Heavy Flavor Tracker Extend the measurement capabilities in the heavy flavor domain, good probe to QGP: • Direct topological reconstruction of charm hadrons (e.g. D0 → K π, cτ ∼ 120 µm)
The STAR detector
HFT
@ RHIC
3
200 GeV Au-Au collisions @ RHIC dNch/dη ~ 700 in central events
TPC – Time Projection Chamber (main tracking detector in STAR)
HFT – Heavy Flavor Tracker
SSD – Silicon Strip Detector IST – Intermediate Silicon Tracker PXL – Pixel Detector
σ = ~1 mm
σ = ~300 µm
σ = ~250 µm
σ = <30 µm
Tracking inwards with gradually improved resolution:
R (cm)
Need to resolve displaced vertices in high multiplicity environment
G. Contin | [email protected]
The PiXeL detector (PXL)
PIXEL2016 - Sestri Levante 4
First vertex detector at a collider experiment based on Monolithic Active Pixel Sensor technology
9/5/2016
G. Contin | [email protected]
PXL Design Parameters
PIXEL2016 - Sestri Levante 5
DCA Pointing resolution (10 ⊕ 24 GeV/p⋅c) µm
Layers Layer 1 at 2.8 cm radius Layer 2 at 8 cm radius
Pixel size 20.7 µm X 20.7 µm
Hit resolution 3.7 µm (6 µm geometric)
Position stability 5 µm rms (20 µm envelope)
Material budget first layer X/X0 = 0.39% (Al conductor cable)
Number of pixels 356 M
Integration time (affects pileup) 185.6 µs
Radiation environment 20 to 90 kRad / year 2*1011 to 1012 1MeV n eq/cm2
Rapid detector replacement < 1 day
356 M pixels on ~0.16 m2 of Silicon
9/5/2016
G. Contin | [email protected]
Material budget on the inner layer Thinned sensor: 50 µm - 0.068% X0 Aluminum-conductor cable (32 µm-thick traces) - 0.128% X0 Carbon fiber stiffener (125 µm) and sector tube (250 µm) - 0.193% X0 Air cooled
Mechanical support with kinematic mounts (insertion side)
Cantilevered support
PXL System Overview 10 sectors total 5 sectors / half
4 ladders / sector 10 sensors / ladder
carbon fiber sector tubes
Highly parallel system
PIXEL2016 - Sestri Levante 6 9/5/2016
Ladder with 10 MAPS sensors (~ 2×2 cm each)
0.388% X0
G. Contin | [email protected]
PXL MAPS sensor
4 sub-arrays to help with process variation
Digital section End-of-column discriminators Integrated zero suppression (up to 9 hits/row) Ping-pong memory for frame readout (~1500 w) 2 LVDS data outputs @ 160 MHz
Ultimate-2: third MIMOSA-family sensor revision developed for PXL at IPHC, Strasbourg
Monolithic Active Pixel Sensor technology
Pixel matrix 928 rows * 960 columns = ~1M pixel In-pixel amplifier In- pixel Correlated Double Sampling (CDS)
PIXEL2016 - Sestri Levante 7
High resistivity p-epi layer Reduced charge collection time Improved radiation hardness
S/N ~ 30 MIP Signal ~ 1000 e- Rolling-shutter readout A row is selected For each column, a pixel is
connected to discriminator Discriminator detects possible hit Move to next row
185.6 µs integration time ~170 mW/cm² power dissipation
9/5/2016
G. Contin | [email protected]
2 m (42 AWG TP)
11 m (24 AWG TP)
100 m (fiber optic)
Highly parallel system 4 ladders per sector 1 Mass Termination Board (MTB) per sector 1 sector per RDO board 10 RDO boards in the PXL system
RDO motherboard w/ Xilinx Virtex-6 FPGA
DAQ PC with fiber link to RDO board
Mass Termination Board (signal buffering) + latch-up protected power
PXL Detector Powering and Readout Chain
PIXEL2016 - Sestri Levante 8
Clk, config, data, power
Clk, config, data
PXL built events
Trigger, Slow control, Configuration, etc.
Existing STAR infrastructure
PXL Sector
9/5/2016
G. Contin | [email protected]
10 PIXEL2016 - Sestri Levante
PXL Production, Assembly and QA
Production goals: • 2 detector copies • 40 additional spare
ladders for refurbishment • 120 high quality ladders
Production processes: • Sensor probe testing • Ladder assembly • Ladder testing • Sector assembly • Detector-half assembly
G. Contin | [email protected]
PXL Probe Testing Thinned and diced 50 µm-thick sensors Full sensor characterization Full speed readout (160 MHz)
Custom made vacuum chuck Testing up to 18 sensors per batch
(optimized for sensor handling in 9-sensor carrier boxes)
Manual alignment (~1 hr) LabWindows GUI for system control Automated interface to a database
Probe card with readout electronics
Sensors built-in testing functionality Proper probe pin design for
curved thinned sensors Yield 46% - 60% (spare probe cards) Administrative control of sensor ID
curved thinned sensors
11 9/5/2016 PIXEL2016 - Sestri Levante
G. Contin | [email protected]
Ladder Assembly
12 PIXEL2016 - Sestri Levante 9/5/2016
Precision vacuum chuck fixtures to position sensors by hand Sensors are positioned with butted edges. Acrylic adhesive prevents CTE difference based damage. Weights taken at all assembly steps to track material and as QA.
Assembled ladder
Cable reference holes for assembly
Hybrid cable with carbon fiber stiffener plate on back in position to glue on sensors.
Sensor positioning
FR-4 Handler
G. Contin | [email protected]
13
From ladders to sectors… to detector halves
Metrology picture
Sectors Ladders are glued on carbon fiber sector tubes in 4 steps Pixel positions are measured and related to tooling balls After touch probe measurements, sectors are tested
electrically for damage from metrology
Detector half Sectors are mounted in dovetail slots on
detector half Metrology is done to relate sector tooling
balls to each other and to kinematic mounts
Sector assembly fixture
A detector half
Sector in the metrology setup
9/5/2016 PIXEL2016 - Sestri Levante
G. Contin | [email protected]
PXL Position Control
Position stability Vibration at air cooling full flow: ~5 µm RMS Stable displacement at full air flow: ~30 µm Stable displacement at power on: ~5 µm
Global hit position resolution: ~ 6.2 µm
14 PIXEL2016 - Sestri Levante
Vibration test through capacitive probe
Metrology survey 3D pixel positions on sector (within ~10 µm) are
measured with touch probe and related to tooling balls Sector tooling ball positions related to kinematic
mounts to relate pixel positions to final PXL location
Detector-half is fully mapped
HFT DCA pointing resolution: ( 10 ⊕ 24/p) µm
9/5/2016
Sector in the metrology setup
G. Contin | [email protected]
PXL Installation in STAR
PIXEL2016 - Sestri Levante 15
Kinematic mounts Novel insertion approach Inserted along rails and
locked into a kinematic mount inside the support structure
It can be replaced in < 1 day with a spare copy
9/5/2016
duplicate, truncated PXL support tube with kinematic mounts
Insertion in STAR PXL inserted and cabled in STAR
G. Contin | [email protected]
Overall stats #
Assembled ladders 146
Installed on sectors 127
Ladder tests ~2000
-5
0
5
10
15
20
25
30
2013
---
- Jan Feb
Mar
Apr
May Jun Jul
Aug Sep
Oct
Nov
Dec
2014
---
- Jan Feb
Mar
Apr
May Jun Jul
Aug Sep
Oct
Nov
Dec
2015
---
- Jan Feb
Mar
Apr
May Jun Jul
Aug Sep
Oct
Nov
Dec
Ladder and Sector construction Ladders on sector
Ladders assembled
PXL production timeline
Primary Detector construction
Spare Detector construction
Al Inner Ladders + Run14 refurbishment
Run15 refurbishment
Delivered: 2014 Run: Primary Detector, 2 aluminum cable inner ladders, others in copper 2015-2016 Runs: 2 detector copies, all inner ladders in aluminum
Sector refurbishment:
After each STAR Run for latch-up induced damage After power supply accident during 2015 Run installation
9/5/2016 PIXEL2016 - Sestri Levante 16
G. Contin | [email protected]
iWoRID '14 - Trieste 17
Operations, Performance, Lessons Learned
6/23/2014
G. Contin | [email protected]
2013 Engineering Run
Sensor IR picture Flawed ladder dissection: searching for shorts
Inner layer design
PXL Engineering Run assembly crucial to deal with a number of unexpected issues
Limited capability to remotely control power and current limits After the engineering run added functionality to the Mass Termination Board:
remote setting of LU threshold and ladder power supply voltage + current and voltage monitoring
Shorts between power and gnd, or LVDS outputs
Adhesive layer extended in both dimensions to increase the portion coming out from underneath the sensors
Insulating solder mask added to low mass cables
Mechanical interference in the driver boards on the existing design.
The sector tube and inner ladder driver board have been redesigned to give a reasonable clearance fit
Inner layer design modification: ~ 2.8 cm inner radius
Engineering run geometry
18 PIXEL2016 - Sestri Levante 9/5/2016
G. Contin | [email protected]
PXL data taking Timeline:
PXL Operations
Hit multiplicity per sensor: up to 1000/inner-sensor, 100/outer-sensor Dead time up to ~6% Typical trigger rate: 0.8-1 kHz Latch-up reset events: 2 latch-up/min Periodic reset to clear SEUs
Collected minimum bias events in the PXL acceptance:
2014 Run: ~ 1.2 Billion Au+Au @ √sNN = 200GeV
2015 Run: @ √sNN = 200GeV
2016 Run: @ √sNN = 200GeV
~ 1 Billion p+p ~ 0.6 Billion p+Au ~ 2 Billion Au+Au ~ 0.3 Billion d+Au
19 PIXEL2016 - Sestri Levante 9/5/2016
2013 2014 2015 2016
2014-2016: HFT data taking PXL Engineering Run HFT decommissioned
PXL R&D
~ 2003
G. Contin | [email protected]
Operational issues: Latch-up damage
Unexpected damage seen on 15 ladders in the STAR radiation environment in 2014 Run first 2 weeks Digital power current increase Sensor data corruption Hotspots in sensor digital section Correlated with latch-up events Limited with operational methods
Latch-up tests at BASE facility (LBL) to measure latch-up
cross-section and reproduce damage 50 µm & 700 µm thick, low and high resistivity sensors; PXL ladders Irradiation with heavy-ions and protons
Results and observations
Current limited latch-up states observed (typically ~300 mA) Damage reproduced only with HI on PXL 50 µm thinned sensors
Safe operations envelope implemented
Latch-up protection at 80 mA above operating current Periodic detector reset to clear SEU
Latch-up phenomenon: • Self feeding short circuit
caused by single event upset • Can only be stopped by
removing the power
Inner layer damage: 14%
PIXEL2016 - Sestri Levante 20
Operating current evolution. Threshold is set at ~1.2A here
9/5/2016
G. Contin | [email protected]
Latch-up test setup and damage analysis
PIXEL2016 - Sestri Levante 21
Individual sensor test boards and ladders mounted on cooling plate
6 hot spots this location 5 hot spots this location
IR “hotspots” locating the damage tend to favor particular structures (isolated buffers with specific structure pitch)
Pixel sensor layers deconstructed (plasma etching technique) and viewed with SEM. The layers appear to be melted
1 µm
9/5/2016
G. Contin | [email protected]
Damage evolution
PIXEL2016 - Sestri Levante 22
Run Good sensors on Inner Layer
Good sensors on Outer Layer Comment
installation end of run installation end of run
2014 100% 82% 100% 95% LU damage, most of it in the first 15 days of operations
2015 99% 94% 98% 96% (93%)* * = Lost control of an outer ladder (10 good sensors off)
2016 100% 95% (87%)+ 99% 98% + = Current instability on inner ladder (8 good sensors off)
9/5/2016
Good sensor = sensor with >95% active channels and uniform efficiency
2014 2015 2016
G. Contin | [email protected]
DCA pointing resolution Design requirement exceeded: 46 µm for 750 MeV/c
Kaons for the 2 sectors equipped with aluminum cables on inner layer
~ 30 µm for p > 1 GeV/c
From 2015: all sectors equipped with aluminum cables on the inner layer
Physics of D-meson productions High significance signal Nuclear modification factor RAA
Collective flow v2
HFT Performance from 2014 data
PIXEL2016 - Sestri Levante 23
D0 → K π production in √sNN = 200GeV Au+Au collisions (First reconstruction of a small subsample)
46 µm
9/5/2016
G. Contin | [email protected]
Software/Firmware issues
9/5/2016 PIXEL2016 - Sestri Levante 24
Event decoder issue in 2014 data reconstruction A bug in the PXL hit decoding software led to an efficiency loss in the reconstructed
2014 Run data, affecting the preliminary STAR results
Readout firmware issue - 2015 efficiency loss A subtle bug introduced by a change in the PXL RDO firmware led to an efficiency
loss in the 2015 Run data The extensive tests with pattern data and the performance of full detector calibrations
were inadequate to find this problem A fast-offline tracking QA was put in place only after the 2015 Run A post-run investigation based on external sensor illumination with LED allowed for
firmware debugging and correction
After fixing the bug, the new data reconstruction and analysis showed a significant improvement in the performance, which now matches the simulation
data - simulation comparison
D0 → K π significance improvement
G. Contin | [email protected]
Conclusions
9/5/2016 PIXEL2016 - Sestri Levante 25
The first generation MAPS-based detector at a collider experiment successfully completed the 3-year physics program at RHIC
As part of the Heavy Flavor Tracker, the PXL detector enabled STAR to perform a direct topological reconstruction of the charmed hadrons
The 2013 Engineering Run was crucial for dealing with the unexpected problems that developed during the following physics data taking
Due to beam-induced damage, the PXL construction phase continued throughout the entire detector life for yearly refurbishment and optimization
The relatively short duration of the HFT program is not optimal to exploit such a complex detector system, nevertheless the project was successfully completed